一种基于神经网络的纳秒脉冲波形重建方法  

A nano-second pulse waveform reconstruction method based on neural network

作　　者：吕东辉 程杰[1,2] 李锐 张楠[1,2] 张立刚 LüDonghui;Cheng Jie;Li Rui;Zhang Nan;Zhang Ligang(Northwest Institute of Nuclear Technology,Xi’an 710024,China;Key Laboratory of Advanced Science and Technology on High Power Microwave,Xi’an 710024,China)

机构地区：[1]西北核技术研究所,西安710024 [2]先进高功率微波技术重点实验室,西安710024

出　　处：《强激光与粒子束》2025年第1期33-38,共6页High Power Laser and Particle Beams

基　　金：先进高功率微波技术重点实验室课题。

摘　　要：针对一种由高速数采通道存在寄生参数、带宽不足导致的纳秒脉冲测量波形畸变的问题,提出了一种基于神经网络的波形重建方法。通过单一神经网络辨识高速数采畸变波形与示波器参考波形的局部映射关系,通过神经网络序列完成全局波形的重建。验证实验表明所提出的方法可以明显缓解高速数采波形的边沿变缓、过冲等问题,波形功率估计精度提高32.5%,能够显著改善高速数采的频响特性。A new method of waveform reconstruction based on neural network is proposed to solve the problem of nano-second pulse distortion,which is caused by the existence of parasitic parameters and insufficient bandwidth in high-speed digital acquisition channels.The local mapping relationship between the distortion waveform acquired by the high-speed digital acquisition system and the reference waveform obtained from the oscilloscope is identified through single neural networks.Then,the global waveform is reconstructed by a series of neural networks.The experimental results show that the proposed method can obviously alleviate the problems such as the edge delay,overshoot of the distortion waveform,thus it can improve the power estimation accuracy by 32.5%,as well as improve the frequency response characteristics of the high-speed digital acquisition system.

关 键 词：神经网络 波形重建 纳秒脉冲 波形畸变 频响特性 

分 类 号：TM836[电气工程—高电压与绝缘技术]